tejohnson (Teresa Johnson)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 27 2015, 11:17 AM (142 w, 2 d)

Recent Activity

Yesterday

tejohnson added a comment to D42107: [ThinLTO] - Stop internalizing non-prevailing symbols..

@tejohnson Can't we use ThinLTO.PrevailingModuleForGUID in the backend?

Wed, Jan 17, 12:06 PM
tejohnson added a comment to D42107: [ThinLTO] - Stop internalizing non-prevailing symbols..

@grimar Can't internalization be skipped during backend processing rather than changing linkage in summary index?. At the very first glance it is possible to check if def is prevailing or not in MustPreserveGV (see thinLTOInternalizeModule)

Wed, Jan 17, 11:48 AM

Fri, Jan 12

tejohnson added a comment to D41297: [ThinLTO] Implement summary visualizer.

At which point of the API is this guarantee?

  1. ThinLTO.ModuleMap stores BitcodeModule objects and uses them during entire thin link phase (until backend threads are launched)
  2. BitcodeModule contains pointers to BC data, so this data should be actual for module to be parsed)
  3. Module summary is exported to .dot in CombinedIndexHook which is invoked in the beginning of thin link phase

None of these seems like API guarantee to me, it sounds like "this is what actually happens in practice today in this particular workflow".

Fri, Jan 12, 11:18 AM
tejohnson added inline comments to D41297: [ThinLTO] Implement summary visualizer.
Fri, Jan 12, 10:31 AM
tejohnson added a comment to D41297: [ThinLTO] Implement summary visualizer.

@tejohnson The BM.parseModule is invoked in the very end of thin link (see runThinLTOBackendThread). How can it work if it's not guaranteed that we have valid BC data till this moment? Note: parse module uses Stream

Fri, Jan 12, 9:57 AM
tejohnson added a comment to D41297: [ThinLTO] Implement summary visualizer.

At which point of the API is this guarantee?

  1. ThinLTO.ModuleMap stores BitcodeModule objects and uses them during entire thin link phase (until backend threads are launched)
  2. BitcodeModule contains pointers to BC data, so this data should be actual for module to be parsed)
  3. Module summary is exported to .dot in CombinedIndexHook which is invoked in the beginning of thin link phase
Fri, Jan 12, 9:42 AM
tejohnson added a comment to D41988: [ThinLTO] - Set WeakAnyLinkage for all LinkerRedefined symbols..

The same handling occurs for regular LTO (setting LinkerRedefined to WeakAny only when prevailing). Is this a problem there? If not, why?

Yeah, I saw the same code in regular LTO,
but it is not a problem for it, our defsym.ll testcase mentioned has both ThinLTO and regular LTO sub-cases and that problem occurs only for ThinLTO case.

Answering my own question perhaps, but I guess this is because in the ThinLTO case we end up importing bar2 and therefore promoting it to external linkage instead of leaving it alone which would cause it to be internalized?

That part I believe is exactly what I observed during debugging of ThinLTO.

Fri, Jan 12, 8:24 AM
tejohnson added a comment to D41988: [ThinLTO] - Set WeakAnyLinkage for all LinkerRedefined symbols..

The same handling occurs for regular LTO (setting LinkerRedefined to WeakAny only when prevailing). Is this a problem there? If not, why? Answering my own question perhaps, but I guess this is because in the ThinLTO case we end up importing bar2 and therefore promoting it to external linkage instead of leaving it alone which would cause it to be internalized?

Fri, Jan 12, 7:53 AM
tejohnson added a comment to D41297: [ThinLTO] Implement summary visualizer.

Sorry for the delay! I have one big concern below. Will look at the code that does the graph emission next, I've only skimmed that part.

Fri, Jan 12, 7:11 AM

Tue, Jan 9

tejohnson committed rL322103: Fix crash when linking metadata with ODR type uniquing.
Fix crash when linking metadata with ODR type uniquing
Tue, Jan 9, 10:34 AM
tejohnson closed D41624: Fix crash when linking metadata with ODR type uniquing.
Tue, Jan 9, 10:34 AM
tejohnson added a comment to D41624: Fix crash when linking metadata with ODR type uniquing.

The test still seems complex to me (do you need all the IR types? The code in the functions? And all these metadatas and metadata fields?

Tue, Jan 9, 10:00 AM

Mon, Jan 8

tejohnson added a comment to D41624: Fix crash when linking metadata with ODR type uniquing.

Ping

Mon, Jan 8, 2:21 PM
tejohnson added a comment to D39869: [Inliner] Inline through indirect call sites having !callees metadata.

Ping

Mon, Jan 8, 2:20 PM
tejohnson added a comment to D39869: [Inliner] Inline through indirect call sites having !callees metadata.

Ping

Mon, Jan 8, 2:16 PM

Tue, Jan 2

tejohnson created D41669: Use ODR debug type uniquing when enabled during function cloning.
Tue, Jan 2, 10:35 AM

Fri, Dec 29

tejohnson updated the diff for D41624: Fix crash when linking metadata with ODR type uniquing.

Address comments.

Fri, Dec 29, 7:55 AM
tejohnson added inline comments to D41624: Fix crash when linking metadata with ODR type uniquing.
Fri, Dec 29, 7:54 AM

Thu, Dec 28

tejohnson created D41624: Fix crash when linking metadata with ODR type uniquing.
Thu, Dec 28, 9:50 PM

Fri, Dec 22

tejohnson accepted D41497: [ThinLTO][CachePruning] explicitly disable pruning.

LGTM assuming the answer to my question below is "yes". =)

Fri, Dec 22, 10:18 AM
tejohnson added a comment to D41497: [ThinLTO][CachePruning] explicitly disable pruning.

Thanks this looks much nicer. Question below about the handling of the Optional though.

As an aside, it would be great to have similar handling in parseCachePruningPolicy(), which is used by gold-plugin and lld, and which gives an error when -1 is passed as the interval. But that doesn't need to be in this patch, just noting it here.

If possible I would rather do that in another change.

Fri, Dec 22, 8:37 AM
tejohnson added a comment to D41497: [ThinLTO][CachePruning] explicitly disable pruning.

Thanks this looks much nicer. Question below about the handling of the Optional though.

Fri, Dec 22, 7:29 AM
tejohnson accepted D41489: [ThinLTO] Don't try to import noinline functions.

LGTM with one minor comment nit.
Thanks!

Fri, Dec 22, 6:41 AM

Thu, Dec 21

tejohnson added a comment to D37993: [ThinLTO/gold] Implement ThinLTO cache pruning support.
In D37993#893466, @pcc wrote:

Is it a good idea to support cache pruning in the gold plugin? As I mentioned in D31063 there are some unavoidable race conditions due to limitations in the plugin API.

Gold plugin

In D37993#918921, @pcc wrote:
In D37993#912428, @pcc wrote:
In D37993#893466, @pcc wrote:

Is it a good idea to support cache pruning in the gold plugin? As I mentioned in D31063 there are some unavoidable race conditions due to limitations in the plugin API.

Ping... not sure whether it's a good idea to ship a known racy feature.

If you really want this feature in gold I'd suggest extending the gold plugin API so that it accepts file descriptors, and then have the plugin support this feature conditionally on whether the gold version is new enough.

Hi Peter, thanks for pointing this out, and sorry for the delay in responding. Wondering whether instead of removing cache pruning support from gold-plugin for now, we could do something else: write the MemoryBuffer to another temp file and use that path (setting IsTemporary to true so it gets cleaned up). This would obviously require more space on disk, but maybe worth it to enable the pruning safely.

Copying to a temporary file sounds fine to me. And if gold ever gets support for file descriptors we could always use the copying code path as a fallback for older versions.

Yi, can you address these suggestions?

Sure, will fix these in a follow up change.

Thu, Dec 21, 10:25 AM
tejohnson added inline comments to D41474: Fix a crash in lazy loading of Metadata in ThinLTO.
Thu, Dec 21, 9:09 AM
tejohnson added a reviewer for D41474: Fix a crash in lazy loading of Metadata in ThinLTO: mehdi_amini.
Thu, Dec 21, 9:09 AM
tejohnson added inline comments to D41340: [LTO] Remove temporary files corresponding to stream which LTO skips.
Thu, Dec 21, 8:10 AM
tejohnson added a comment to D41489: [ThinLTO] Don't try to import noinline functions.

Thanks! Couple suggestions.

Thu, Dec 21, 8:03 AM

Wed, Dec 20

tejohnson committed rL321197: [PGO] Function section hotness prefix should look at all blocks.
[PGO] Function section hotness prefix should look at all blocks
Wed, Dec 20, 9:54 AM
tejohnson closed D41395: [PGO] Function section hotness prefix should look at all blocks.
Wed, Dec 20, 9:54 AM

Tue, Dec 19

tejohnson created D41395: [PGO] Function section hotness prefix should look at all blocks.
Tue, Dec 19, 7:24 AM

Dec 18 2017

tejohnson committed rL321018: [PGO] Fix handling of cold entry count for instrumented PGO.
[PGO] Fix handling of cold entry count for instrumented PGO
Dec 18 2017, 12:03 PM
tejohnson closed D41307: [PGO] Fix handling of cold entry count for instrumented PGO.
Dec 18 2017, 12:03 PM
tejohnson committed rL321000: [ThinLTO] Make distributed indexes test more robust.
[ThinLTO] Make distributed indexes test more robust
Dec 18 2017, 10:01 AM
tejohnson added a comment to D41231: [Support][CachePruning] Fix regression that prevents disabling of pruning.

Just a couple of minor comments - thanks for adding the test. With those fixed it looks good from my perspective, but please wait for labath to sign off.

Dec 18 2017, 9:48 AM

Dec 15 2017

tejohnson committed rL320903: Add another missing -enable-import-metadata to test.
Add another missing -enable-import-metadata to test
Dec 15 2017, 5:36 PM
tejohnson committed rL320899: Add -enable-import-metadata to test.
Add -enable-import-metadata to test
Dec 15 2017, 5:01 PM
tejohnson committed rL320897: Fix NDEBUG build problem in r320895.
Fix NDEBUG build problem in r320895
Dec 15 2017, 4:30 PM
tejohnson committed rL320895: [ThinLTO] Enable importing of aliases as copy of aliasee.
[ThinLTO] Enable importing of aliases as copy of aliasee
Dec 15 2017, 4:19 PM
tejohnson closed D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.
Dec 15 2017, 4:19 PM
tejohnson updated the diff for D41307: [PGO] Fix handling of cold entry count for instrumented PGO.

Address comment

Dec 15 2017, 1:46 PM
tejohnson created D41307: [PGO] Fix handling of cold entry count for instrumented PGO.
Dec 15 2017, 1:09 PM
tejohnson accepted D41279: [ThinLTO][C-API] Correct api comments.

LGTM

Dec 15 2017, 12:57 PM
tejohnson updated the diff for D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.

Address comments

Dec 15 2017, 12:00 PM
tejohnson added a comment to D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.
In D40747#956983, @pcc wrote:
In D40747#947144, @pcc wrote:

Now that we basically treat alias summaries in the same way as their aliasees, I wonder whether we can simplify things by removing AliasSummary and summarizing aliases by creating a FunctionSummary or GlobalVariableSummary from the aliasee instead.

I'm not sure we can as there are other requirements for knowing the alias relationship. For example, we can't convert a non-prevailing alias and aliasee to available_externally (see thinLTOResolveWeakForLinkerGUID) - that is in the original module and not related to importing.

That could be resolved by adding a HasAlias flag to GlobalValueSummary, no? We would set it on aliases and anything that is aliased.

Dec 15 2017, 12:00 PM
tejohnson added a comment to D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.

ping - Peter, would you be able to take a look at the changes?

Dec 15 2017, 10:03 AM
tejohnson added a comment to D41297: [ThinLTO] Implement summary visualizer.

Nice! This looks very useful. What exactly does the python script do? I suppose the script must have been used to create the attached .png graph, since it has GV names and not GUIDs?

Dec 15 2017, 9:25 AM
tejohnson added a comment to D41231: [Support][CachePruning] Fix regression that prevents disabling of pruning.

Add a test? You can use llvm-lto to trigger this legacy interface - although it looks like no support was added to llvm-lto to invoke this interface. It should be straightforward to add that (see where it calls setCacheDir).

I will attempt to do this.

Are you aware of any users of the legacy C++ api - i.e the one exposed by lvm-lto?

Dec 15 2017, 9:03 AM
tejohnson added a comment to D41231: [Support][CachePruning] Fix regression that prevents disabling of pruning.

Add a test? You can use llvm-lto to trigger this legacy interface - although it looks like no support was added to llvm-lto to invoke this interface. It should be straightforward to add that (see where it calls setCacheDir).

Dec 15 2017, 8:20 AM
tejohnson added inline comments to D41279: [ThinLTO][C-API] Correct api comments.
Dec 15 2017, 8:16 AM
tejohnson added inline comments to D41279: [ThinLTO][C-API] Correct api comments.
Dec 15 2017, 8:15 AM
tejohnson accepted D41291: [ThinLTO] Disallow multiple prevailing defs.

LGTM

Dec 15 2017, 7:47 AM
tejohnson updated subscribers of D41291: [ThinLTO] Disallow multiple prevailing defs.
Dec 15 2017, 7:47 AM
tejohnson added inline comments to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.
Dec 15 2017, 6:52 AM

Dec 14 2017

tejohnson added a comment to D41222: Handle previously ASAN-instrumented IR gracefully when ASAN re-invoked.
In D41222#955655, @pcc wrote:

If this doesn't even fix the ThinLTO bug I don't see why we should land it. The other scenarios mentioned on the bug are "no warranty" scenarios because they use flags like -emit-llvm which aren't really user facing.

Dec 14 2017, 12:21 PM
tejohnson added a comment to D41222: Handle previously ASAN-instrumented IR gracefully when ASAN re-invoked.

Could you please clang-format the patch for consistency?

Dec 14 2017, 10:48 AM
tejohnson updated the diff for D41222: Handle previously ASAN-instrumented IR gracefully when ASAN re-invoked.

Simplify test case

Dec 14 2017, 8:33 AM
tejohnson added a comment to D41222: Handle previously ASAN-instrumented IR gracefully when ASAN re-invoked.
In D41222#954847, @vsk wrote:

Is it possible to reduce the test case s.t the test file doesn't contain instrumented IR? ASan instrumentation is prone to change, so I worry that the test could start passing for invalid reasons. This might happen if 1) __asan_before_dynamic_init is renamed, 2) GlobalsMetadata::doInit is written more defensively, and 3) a non-idempotent ASan transform is introduced later.

I envision something like:

RUN: opt -asan -asan-module < %s -S -o %t.ll
RUN: opt -asan -asan-module < %t.ll -S -o %t.2.ll
RUN: diff %t.ll %t.2.ll

I spent a few minutes trying to come up with a test like this but did not succeed, because my test file didn't trip the assert you hit.

Dec 14 2017, 8:31 AM
tejohnson accepted D41238: [LLVMgold] Use platform dependent extension in tests.

LGTM

Dec 14 2017, 7:52 AM

Dec 13 2017

tejohnson created D41222: Handle previously ASAN-instrumented IR gracefully when ASAN re-invoked.
Dec 13 2017, 7:48 PM
tejohnson added a comment to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.
In D41113#954157, @pcc wrote:

This seems fine to me with the assert.

There's even a probable current issue - if the prevailing def in the asm is marked dead, anything reachable from it might be marked dead as well (i.e. if the only use is in that def).

Module asm definitions do not refer to anything at the summary level, so it does not matter if we mark them as dead. Any symbols that are referenced from module inline asm are GC roots because of D32544.

Dec 13 2017, 12:41 PM
tejohnson added a comment to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.
I meant that the asm_undefined2.ll test should cause this assert without the fix in this patch

Without modification to gold-plugin.cpp assertion fires.

Dec 13 2017, 9:52 AM
tejohnson added a comment to D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.
In D40747#947144, @pcc wrote:

Now that we basically treat alias summaries in the same way as their aliasees, I wonder whether we can simplify things by removing AliasSummary and summarizing aliases by creating a FunctionSummary or GlobalVariableSummary from the aliasee instead.

Dec 13 2017, 8:53 AM
tejohnson added a comment to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.

Can you add an assert for this

check-all passes with this, but we can explicitly mark asm undefined symbol as prevailing using llvm-lto(2) and trigger this assert. Is this OK?

Dec 13 2017, 8:35 AM
tejohnson added a comment to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.

Thanks for sharing this! Here are results of my brief investigation:

IR symbol loading (through IRSymtab) happens absolutely identically in case of gold plugin and lld. In both cases undefined asm counterpart of 'foo' gets FB_used bit (so isUsed() returns true).
The difference, however, shows up when we process symbol resolutions in lto::addModuleToGlobalRes. The irsymtab::storage::Symbol has two name fields, used for different purposes

  • Name. This is mangled name used to map GlobalResolution to specific IR symbol (see lto::addModuleToGlobalRes)
  • IRName. This is unmangled name used to compute GUID (hash). For asm undefined symbols IRName is an empty string as they aren't IR symbols. This is being done in irsymtab::Builder::addSymbol

    Now to your question (why is the handling added in D32544 not covering this?):

    The lto::addModuleToGlobalRes relies on correct setting of 'Prevailing' in SymbolResolution. The logic behind it assumes that prevailing symbol always has valid IRName, so there are following lines of code:

    ` if (Res.Prevailing) GlobalRes.IRName = Sym.getIRName(); ` When we see asm undefined symbol with Prevailing == true (gold plugin) we set GlobalRes.IRName to an empty string. After that we are unable to correctly compute hash and set its IR counterpart as GC root. This in turn causes incorrect symbol internalization/DCE.
Dec 13 2017, 7:35 AM
tejohnson added a reviewer for D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing': pcc.
Dec 13 2017, 7:35 AM

Dec 12 2017

tejohnson added inline comments to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.
Dec 12 2017, 9:09 AM
tejohnson added inline comments to D41113: [LLVMgold] Don't set resolutions for undefined symbols to 'Prevailing'.
Dec 12 2017, 7:50 AM

Dec 11 2017

tejohnson accepted D40970: [ThinLTO] Remove unused (?) code from thinLTOInternalizeModule.

@tejohnson Any comments or objections?

Dec 11 2017, 8:50 AM

Dec 6 2017

tejohnson added a comment to D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.

Just a quick note:

Could you check what the debug info looks like for an imported alias that is inlined, and one that isn't inlined?

Specifically my concern is over whether or not this causes a second CU to be created - currentyl we're threading a needle to avoid fixing/updating GDB by not needing multiple CUs per DWO file due to importing cross-CU inlining into the calling CU, etc.

Can talk more offline or through this thread.

Dec 6 2017, 12:53 PM

Dec 1 2017

tejohnson created D40748: [ThinLTO] Enable importing of aliases as copy of aliasee.
Dec 1 2017, 1:04 PM
tejohnson added a dependent revision for D40747: [ThinLTO] Enable importing of aliases as copy of aliasee: D40748: [ThinLTO] Enable importing of aliases as copy of aliasee.
Dec 1 2017, 1:04 PM
tejohnson created D40747: [ThinLTO] Enable importing of aliases as copy of aliasee.
Dec 1 2017, 1:03 PM

Nov 30 2017

tejohnson accepted D40593: ThinLTOBitcodeWriter: Try harder to discard unused references to the merged module..

LGTM - I assume the modified test fails without the patch? I can't tell from looking at it whether it anon would have dead constant users.

Nov 30 2017, 2:24 PM

Nov 22 2017

tejohnson accepted D40327: CachePruning: Allow limiting the number of files in the cache directory..

LGTM

Nov 22 2017, 10:22 AM
tejohnson added inline comments to D40327: CachePruning: Allow limiting the number of files in the cache directory..
Nov 22 2017, 9:02 AM

Nov 20 2017

tejohnson committed rL318677: [SROA] Correctly invalidate analyses when dead instructions deleted.
[SROA] Correctly invalidate analyses when dead instructions deleted
Nov 20 2017, 10:36 AM
tejohnson closed D39921: [SROA] Correctly invalidate analyses when dead instructions deleted by committing rL318677: [SROA] Correctly invalidate analyses when dead instructions deleted.
Nov 20 2017, 10:36 AM
tejohnson updated the diff for D39921: [SROA] Correctly invalidate analyses when dead instructions deleted.

Remove ModuleID and source_filename

Nov 20 2017, 10:33 AM
tejohnson updated the diff for D39921: [SROA] Correctly invalidate analyses when dead instructions deleted.

Simplify the test case as much as possible

Nov 20 2017, 9:54 AM
tejohnson added a comment to D39921: [SROA] Correctly invalidate analyses when dead instructions deleted.

Ping

Nov 20 2017, 6:38 AM

Nov 14 2017

tejohnson accepted D40056: [ThinLTO] Remove too aggressive assertion in building function call graph..

LGTM
Thanks!

Nov 14 2017, 6:36 PM

Nov 13 2017

tejohnson committed rL318042: [ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang.
[ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang
Nov 13 2017, 7:40 AM
tejohnson added a reviewer for D39921: [SROA] Correctly invalidate analyses when dead instructions deleted: chandlerc.
Nov 13 2017, 7:01 AM

Nov 10 2017

tejohnson committed rL317959: Revert "[ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang".
Revert "[ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang"
Nov 10 2017, 5:09 PM
tejohnson committed rL317952: Add x86-registered-target to REQUIRES for new test.
Add x86-registered-target to REQUIRES for new test
Nov 10 2017, 4:07 PM
tejohnson committed rL317951: [ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang.
[ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang
Nov 10 2017, 3:38 PM
tejohnson closed D39923: [ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang by committing rL317951: [ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang.
Nov 10 2017, 3:37 PM
tejohnson created D39923: [ThinLTO] Handle -fdebug-pass-manager for backend invocations via clang.
Nov 10 2017, 2:10 PM
tejohnson created D39921: [SROA] Correctly invalidate analyses when dead instructions deleted.
Nov 10 2017, 1:29 PM
tejohnson created D39901: ASAN lit support for LTO testing and add back ThinLTO test.
Nov 10 2017, 7:42 AM

Nov 9 2017

tejohnson committed rL317823: Revert new ThinLTO ASAN test until lit support added.
Revert new ThinLTO ASAN test until lit support added
Nov 9 2017, 11:27 AM

Nov 8 2017

tejohnson added a comment to D37993: [ThinLTO/gold] Implement ThinLTO cache pruning support.
In D37993#918921, @pcc wrote:
In D37993#912428, @pcc wrote:
In D37993#893466, @pcc wrote:

Is it a good idea to support cache pruning in the gold plugin? As I mentioned in D31063 there are some unavoidable race conditions due to limitations in the plugin API.

Ping... not sure whether it's a good idea to ship a known racy feature.

If you really want this feature in gold I'd suggest extending the gold plugin API so that it accepts file descriptors, and then have the plugin support this feature conditionally on whether the gold version is new enough.

Hi Peter, thanks for pointing this out, and sorry for the delay in responding. Wondering whether instead of removing cache pruning support from gold-plugin for now, we could do something else: write the MemoryBuffer to another temp file and use that path (setting IsTemporary to true so it gets cleaned up). This would obviously require more space on disk, but maybe worth it to enable the pruning safely.

Copying to a temporary file sounds fine to me. And if gold ever gets support for file descriptors we could always use the copying code path as a fallback for older versions.

Nov 8 2017, 10:12 PM
tejohnson committed rL317728: [ThinLTO] New test needs to require LTO.
[ThinLTO] New test needs to require LTO
Nov 8 2017, 1:48 PM
tejohnson committed rL317723: [ThinLTO] Ensure sanitizer passes are run.
[ThinLTO] Ensure sanitizer passes are run
Nov 8 2017, 12:28 PM
tejohnson committed rL317717: Revert "[ThinLTO] Ensure sanitizer passes are run".
Revert "[ThinLTO] Ensure sanitizer passes are run"
Nov 8 2017, 12:09 PM
tejohnson committed rL317715: [ThinLTO] Ensure sanitizer passes are run.
[ThinLTO] Ensure sanitizer passes are run
Nov 8 2017, 11:47 AM
tejohnson closed D39566: [ThinLTO] Ensure sanitizer passes are run by committing rL317715: [ThinLTO] Ensure sanitizer passes are run.
Nov 8 2017, 11:47 AM
tejohnson committed rL317714: [ThinLTO] Ensure sanitizer passes are run.
[ThinLTO] Ensure sanitizer passes are run
Nov 8 2017, 11:46 AM
tejohnson closed D39565: [ThinLTO] Ensure sanitizer passes are run by committing rL317714: [ThinLTO] Ensure sanitizer passes are run.
Nov 8 2017, 11:46 AM